Deep Learning Benchmarks for Multi-Step Photovoltaic Power Forecasting: Comparative Assessment of GRU, LSTM, and BiLSTM Architectures

Accurate photovoltaic (PV) power forecasting is essential for renewable energy integration, dynamic reserve allocation, and generation scheduling. Unpredicted generation ramps induce substantial voltage and frequency deviations on grid-connected distribution networks. This paper provides an objective benchmark among three deep learning architectures: Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Bidirectional LSTM (BiLSTM). Telemetry spanning one year from an operational 175 kW plant in Reggane, Algeria, is evaluated across 72 h horizons (H=288 steps at 15 min resolution) using a Direct Multi-Input Multi-Output framework under strict chronological splitting. The models are systematically evaluated across 1848 rolling 72 h forecast windows using a controlled benchmark protocol under daylight-only conditions (GHI>5 W/m2) against a Diurnal Smart Persistence baseline. Accuracy is assessed via daylight capacity-normalized MAE (nMAE), nRMSE, Diebold–Mariano tests (h=72 lags), and ramp metrics across 34,840 events. The results demonstrate that deep architectures significantly outperform persistence (p<0.001). BiLSTM achieves superior overall accuracy (nMAE=2.84%, nRMSE=3.60%, R2=0.985, skill score = 79.37%) and the highest ramp fidelity (F1-Ramp = 0.965, RME=6.97 kW, yielding 54.95% ramp error reduction). Standard LSTM demonstrates competitive performance (nMAE=3.19%, skill score = 77.95%), whereas GRU exhibits higher deviations (nMAE=5.05%, skill score = 66.91%). Measured sub-millisecond latencies (0.07–0.11 ms/sample) confirm that computational overhead is negligible for 15 min dispatch, establishing BiLSTM as optimal for operational dispatch.